import { NgramMap, TopicToken } from './types';
/**
 * Extracts noun-anchored edge-grams from a token array and accumulates them into `nGrams`.
 *
 * An edge-gram is a contiguous slice of `nGramSize` tokens where:
 * - The **first** and **last** tokens are topic entities (noun or wiki title).
 * - Every token is either a topic entity or a common stop word (e.g. "of", "the").
 * - Every token meets the `minWordLength` character threshold.
 *
 * This allows natural multi-word keyphrases like "state of the art" or
 * "machine learning" while ignoring pure function-word sequences.
 *
 * @param nGramSize - Number of tokens in the slice to evaluate.
 * @param terms - Full token array for the current sentence.
 * @param index - Start position for this slice within `terms`.
 * @param nGrams - Accumulator map mutated in-place: `nGrams[size][phrase] = [sentenceIdx, ...]`.
 * @param minWordLength - Minimum character length for any word to be included.
 * @param sentenceIndex - Index of the originating sentence, appended to the phrase's entry.
 * @returns The same `nGrams` reference (mutated).
 *
 * @example
 * const terms: TopicToken[] = [["machine", 1, 4, ""], ["learning", 1, 5, ""]];
 * const nGrams: NgramMap = {};
 * extractNounEdgeGrams(2, terms, 0, nGrams, 3, 0);
 * // nGrams[2]["machine learning"] === [0]
 */
export declare function extractNounEdgeGrams(nGramSize: number, terms: TopicToken[], index: number, nGrams: NgramMap, minWordLength: number, sentenceIndex: number): NgramMap;